The European Central Bank has published new survey data showing that euro-area companies are taking a sharply different financing path for artificial intelligence than their US counterparts. Based on the ECB’s Survey on the Access to Finance of Enterprises, covering 5,087 firms, 72% of companies planning AI investment expect to use internal funds such as cash flow and retained earnings. Bank loans and grants were each cited by 16%, leasing by 15%, private equity or venture capital by just 6%, and debt securities by only 1%. Another 18% selected no financing option at all.
The investment priorities also go beyond hardware. Some 49% of firms expect to spend on AI technologies and tools, 46% on employee training, 40% on data and infrastructure, and only 12% on hiring AI specialists. The ECB itself warns that this dependence on internal cash could limit how quickly European companies can scale AI adoption, especially for smaller or faster-growing businesses that need large amounts of external capital.
The contrast with the United States is stark. US hyperscalers, data-center developers and infrastructure companies are increasingly funding the AI buildout with borrowed money. A Nikkei study cited by Fortune found that the five largest US tech companies hold $1.65 trillion in hidden debt and $1.35 trillion in on-balance-sheet debt, roughly eight times the level of four years earlier. Moody’s estimates off-balance-sheet deals at $1.2 trillion, with more than $820 billion committed to data centers that are not yet ready. Hyperscalers and related names such as Nvidia have issued $225 billion in bonds in 2026, according to S&P Global, a jump of 973.7% as of mid-year, with projections near $400 billion by year-end. Goldman Sachs expects hyperscaler debt to grow by another 60% in 2027 to a record $420 billion.
Wall Street demand for AI-linked debt is already cooling. Apollo Global’s Torsten Slok reported that investor orders per dollar of hyperscaler bonds fell below two times in July from nearly five times in February. Colby Stilson, head of fixed income at Brown Advisory in London, said the firm is being very selective in hyperscaler debt.
Europe’s financing caution also intersects with a major computing gap. Oxford Economics projects US corporate spending on AI hardware and infrastructure will grow 40% in real terms between 2021 and the end of 2027, while the euro area grows only 12%. Former ECB President Mario Draghi has warned that the European Union hosts under 5% of the world’s AI compute capacity versus 75% for the United States, and that the shortfall between demand and installed supply could widen to 14 gigawatts by 2030. Draghi argued that being cut off from AI could be comparable to being cut off from the US financial system, with catastrophic effects, and proposed that European firms pool buying power to finance new data centers.
